Improved brain storm optimization algorithm for solving multimodal multiobjective problems

Aiming at the problem that multimodal multi-objective optimization is hard to find sufficient equivalent solutions and maintain decision space diversity，a differential brain storm optimization algorithm based on zoning search and non-dominated special crowding distance sort algorithm was proposed．In the proposed algorithm，zoning search divided the decision space into multiple subspaces to reduce search difficulty and maintain population diversity．The k-means clustering strategy could locate and maintain various Pareto optimal solutions，and non-dominated special crowding distance sorting could consider the diversity of decision and objective space and serve as an environmental selection operator to filter solutions．The difference mutation operator replaced the traditional new individual generation operator to enhance the population′s diversity and help locate multiple equivalent optimal solutions．Compared with 5 algorithms，the performance of the zoning search and non-dominated special crowding distance sort algorithm was validated on 13 multimodal multi-objective test functions．Experimental results show that the zoning search and non-dominated special crowding distance sort algorithm performs better than the other 5 algorithms on 11 test functions，and zoning search and non-dominated special crowding distance sort algorithm can find as many equivalent Pareto-optimal sets as possible in the decision space and guarantee a good Pareto front distribution in the objective space. © 2024 Huazhong University of Science and Technology. All rights reserved.